Representation of texts in structured form

Abstract

Although the existing knowledge representation techniques, ranging from the relational databases to the most recent Semantic web languages, are successfully applied in numerous practical applications, they are still unable to represent the information contained in text documents and web pages in structured form, suitable for productive text processing. Text files can represent text documents with no loss of information, however, this information is represented in an unstructured form. Various knowledge formalisms used in different phases of Natural Language Understanding, such as lexical, syntactic, semantic, pragmatic and discourse analysis, are still unable to represent texts in structured form with no loss of information. In this paper, we define the crucial requirements for structured text representation and then, we give a brief introduction to a representation technique that fulfills all these requirements, including the basic data types and learning techniques used to create, maintain and interpret the resulting representation formalism.